At the top of the preprocessing section, you will find two summary tables: the
Cell-level Observations table (1) and the Gene-level Observations table (2).
These tables contain the first quality metrics generated right after uploading your data.
As you move forward through the pipeline, new information will automatically be added to these tables.
For the Cell-level Observations table, you will see details such as the unique identity of each cell (1),
the number of genes detected per cell, the total counts, and other relevant metrics.
As the workflow progresses, any new cell-level metrics will be stored here.
This table can also be downloaded (2), and you can return to the pipeline view at any time by clicking the close button (3).
For the Gene-level Observations table, you will find the gene ID (1),
the total counts per gene, and whether a gene is annotated as mitochondrial, ribosomal, or other categories.
As with the cell table, any new gene-level metrics produced during the workflow will be added here.
This table can also be downloaded (2), and you can return to the pipeline with the close button (3).
During preprocessing, you can filter your dataset according to four parameters:
- Min Genes: (1) filters out low-quality cells with fewer than the indicated number of genes.
- Min Cells: (2) filters out genes expressed in fewer than the indicated number of cells, keeping only biologically relevant genes.
- Mitochondrial Threshold: (3) filters out cells whose mitochondrial percentage is above the indicated value.
- Doublet Detection: (4) enables the detection and removal of doublets that may have been generated during library preparation.
For this tutorial, we will use the default parameters:
200 Min Genes, 3 Min Cells, 5% Mitochondrial Threshold, and Doublet Detection enabled.
Once you set these values, click on Run (5).
What parameters are optimal for preprocessing a single-cell dataset?
Preprocessing of single-cell data is a crucial step for accurate identification and characterization of cell types, states, and biological mechanisms. Thus, proper data preprocessing will improve downstream analysis.
A good starting point to explore the preprocessing parameters is those recommended by Seurat and Scanpy developers such as 3 for Min Genes, 200 for Min Cells, and 5% for Mitochondrial Threshold. However, we encourage users to explore multiple parameters since the best will vary across datasets, species, platforms, etc.
For instance, a systematic analysis determined that a mitochondrial threshold of 5% distinguishes high- and low-quality cells in mouse samples whereas in human failed to discriminate in one third of the datasets.
Droplet-based single-cell methods generates doublets, which are two cells in one droplet. Thus, elimination of doublets is a key step during preprocessing since single-cell methods assumes that each droplet contains only one cell.
After running preprocessing, scExplorer will generate six quality control (QC) plots that help you evaluate
whether the dataset is suitable for downstream analysis:
- Box/Violin plots (top row): show the number of genes per cell, total counts, and the percentage of mitochondrial genes (1)(2)(3).
- Scatter plots (bottom row): show the correlation between the number of genes and total counts,
the relationship between mitochondrial counts and total counts, and the variance of genes across cells (4)(5)(6).
